Virtual Library Generation Enumeration

Pickett et al. [68] describe a program, DIVSEL, for selecting reactants while taking account of the pharmacophoric diversity that exists in the final products. They describe a 2-component library where the reactants in one pool are fixed and a subset of reactants is to be selected from the second pool. The virtual library is enumerated and a pharmacophore key is generated for each of the product molecules. Reactants are selected from the second pool using a dissimilarity-based compound selection process that represents a candidate reactant by a pharmacophore key that covers an ensemble of products. 

Using the above mentioned parameters, we generated a virtual library of enumerated products. 

We calculated a series of physicochemical properties of the virtual libraries generated from enumerations of cores and daughters. Apart from the daughters we derived from the parent compounds, we selected (based on synthetic practicality... 

Once the reactant pools have been filtered, the next step in product-based designs is usually to enumerate the full virtual library. This can be a very time-consuming step and hence a useful precursor can be to enumerate carefully chosen subsets that will give an indication of the success or otherwise of the full virtual experiment. Thus, in a two component reaction it can be useful to take the first reactant in the first pool and combine it with all the reactants in the second pool (to generate 1 x nB products). This should then be followed by the enumeration of one reactant in the second pool with all reactants in the first pool to give nA x 1 products. If either of these two partial enumeration steps fail, then the full enumeration will also fail. Thus, troublesome reactants can be identified early. 

Several product-based approaches to library design that do not require full enumeration have been developed. Pickett et al. have described the design of a diverse amide library where diversity is measured in product space. The DIVSEL program is a DBCS method where dissimilarity is measured in three-point pharmacophore space [83]. Initially, 11 amines were selected based on maximum pharmacophore diversity. Then a total of 1100 carboxylic acids were identified following substructure searching. A set of 1100 pharmacophores keys was generated, where each key corresponds to one acid combined with the 11 amines. DIVSEL was used to select 100 acids based on the diversity of the products. The final library was found to cover 85% of the pharmacophores represented by the entire 12,100 virtual libraries. 

The term virtual library is used quite frequently in the literature these days. Unfortunately there is no common definition on what a virtual library really is. In the simplest case it is a database of fully enumerated structures of products that are under consideration to be made. The generation of such a virtual library involves software that maps the reaction sequence and the corresponding sets of building blocks onto a combinatorial representation (7) (see Note 4). This combinatorial representation can be partially or fully enumerated to generate product structures, which are then stored in structural databases (see Note 5). Subsequently, properties and descriptors to be used in the selection process are to be calculated. As there are limitations in terms of file size etc., such virtual libraries cannot hold more than a few million products. This makes it necessary to filter out reagents (see Subheading 1.3.) prior to construction of the virtual library. 

Several computational methods for generating large databases of chemically reasonable structures (virtual libraries) have been developed. They employ strategies such as the mutation of text strings representing chemical structures, the expansion of Markush structural representations, or virtual combinatorial libraries derived by exhaustive enumeration of all substituent variations at specific points on a core scaffold. An example of these large virtual libraries is the ChemSpace database, containing approx 10 trillion chemical structures for use in similarity and pharmacophore searches, approx 500,000 times more than all the compounds in Chemical Abstracts. 

Reactor can be set up to carry out simple sequential enumeration, combinatorial enumeration, generating combinatorial virtual synthetic libraries. Users also have the option to exclude imwanted products from the enumeration results manually, restricting the outcome of the reaction enumeration process to the desired main products only. Reactor supports the generation of product or reaction libraries in a large variety of different output formats. 

---